[Tarantool-patches] [PATCH 0/2] Validate result of upserts squash & skip invalid upserts which can't be applied

[Vladislav Shpilevoy]

Thanks for the patchset!

I grepped for other vy_apply_upsert() usage cases and found these:

- vy_build_recover_stmt() - should we patch it too?

- vy_tx_write() - here the upsert is skipped in case of any error, but
  what if it was not a ClientError?

On 13/04/2020 23:55, Nikita Pettik wrote:
> Branch: https://github.com/tarantool/tarantool/tree/np/gh-1622-skip-invalid-upserts
> Issue: https://github.com/tarantool/tarantool/issues/1622
> 
> In fact, the first patch in series does not make much sense without
> second since otherwise dump or compaction processes (and any other
> operations that read data as well) will fail whenever they are started.
> But still I've divided fix into two patches for the sake of review
> simplicity.
> 
> Nikita Pettik (2):
>   vinyl: validate resulting tuple after upsert is applied
>   vinyl: skip invalid upserts during squash
> 
>  src/box/vy_history.c                          |  20 ++-
>  src/box/vy_lsm.c                              |  13 +-
>  src/box/vy_tx.c                               |  29 ++--
>  src/box/vy_upsert.c                           |   4 +
>  src/box/vy_write_iterator.c                   |  34 +++--
>  .../vinyl/gh-1622-skip-invalid-upserts.result | 135 ++++++++++++++++++
>  .../gh-1622-skip-invalid-upserts.test.lua     |  50 +++++++
>  7 files changed, 258 insertions(+), 27 deletions(-)
>  create mode 100644 test/vinyl/gh-1622-skip-invalid-upserts.result
>  create mode 100644 test/vinyl/gh-1622-skip-invalid-upserts.test.lua
>